archived-code-advanced

[undone coral]

and actually run the update later, perhaps in LateUpdate in the component

however i am sort of confused what you are trying to do... based on the example how is updating a health bar related to setting the adaptive shake

you can make a fluid API (what this is called) for anything

doesn't mean you should

[flint wraith]

Yeah i did that but it is run before chained functions which already passed values to dotween

[flint wraith]

Well health bar has its settings for example how long it should last, i wanted to add override function that enables it to last for different duration, that call only

[jolly token]

Then you'd make a builder class or struct that wraps around it
You can also always save Tween as variable and modify it

[flint wraith]

but it cant be modified after running also how would struct delay animation start in Update function, after chained functions have been ran

i tried looking dotween code but couldnt find much how they do it

i would says making Update function Coroutine and then waiting 1 frame before calling animation

just to be clear Update function is UpdateHealthBar, not unity Update

[buoyant nacelle]

Can someone help with the code architecture, please? I did a state machine for my actions (attack, move and idle). I have a state manager (this is my context, mono), character base state (abstract class) and classes that inherit from the abstract (move, attack and idle).

Then, I have PlayerInput controller, which has different actions (mouse click, key pressed, etc), and I want this class to switch state. However, I need to pass in the info (for instance, vector3) from PlayerInput to , for example, moving state. What is the best way to implement it? Is it a good architecture?

Another question I have is my moving state have moving logic and animation logic. How should I separate it to follow single responsibility principle? (animation and logic in different classes?)

[lunar jackal]

Why not creating events for this purpose, invoke them in your Input Class and simply add the required callbacks in your actual base class? That way you can also simply add further events for animations and so on
https://learn.microsoft.com/en-us/dotnet/standard/events/

[buoyant nacelle]

that's a good idea. No idea how I missed it. Thanks

[buoyant nacelle]

So, event is invoke in the playerInput and it switches state to let's say move

or abandon state machine all together?

[lunar jackal]

you can still use the state machine, and subscribe to the events send by your player input, but I would try to abstract the input itself as it's pretty nasty to get e.g. cross controller support even with the new input system if you're putting your inputs in the programming

[buoyant nacelle]

i used new input system for that

thanks

[severe topaz]

ia.default.action.performed += callback => yourCurrentState.WhatDo(callback.ReadValue<Vector3>);

or if you're using the component,
OnWhatever( CallbackContext callback)
{ yourCurrentState.WhatDo(callback.ReadValue<Vector3>); }

or something like that, i'm not using the component so i'm not 100% on that

[undone coral]

what does UpdateHealthBar do?

[fair hound]

is there a way to edit instances of a monobehaviour derived script from its inspector reference inside of another script?

[undone coral]

basically postpone running of update until all chained functions are complete
i don't get it, why not just write it that way

write the fluent api so that the last call is update

it doesn't make sense

[flint wraith]

Basically set some sprite colors, transform positions and run dotween with some settings.

I wanted to achieve dotween style of fluent, basically say what you want and then chain the settings. Doing updatehealthbar as last just bit clunkier

Was looking at dotween source code but had enough after about 2 hours 🥹

[undone coral]

okay this is how it looks

class HealthBar : MonoBehaviour {
 Sequence sequence;
 int frame;
 private void SetTween() {
  if (frame != Time.frameCount) {
   frame = Time.frameCount;
   sequence?.Kill();
   sequence = DOTween.Sequence();
  }
 }
 public HealthBar SetShake(...) {
  SetTween();
  sequence.Insert(0, transform.DOShake(...));
  return this;
 }
 public HealthBar UpdateHealthBar(...) {
  ...
 }
 void LateUpdate() {
  // or whatever the method is
  if ((sequence?.IsPlaying() ?? true) == false)) {
   sequence.Play();
  }
 }
}

@flint wraith does that make sense?

there's something here i don't understand because it seems pretty straightforward to me

[flint wraith]

UpdateHealthBar().SetShake() would work?

this is basically dotween style

transform.DoScale(2f, 2f);
or you can do
transform.DoScale(2f, 2f).SetEase(Ease.OutQuart);

in my current model i must do

UpdateHealthBar()
which is ok but if i want to add settings i must do
SetShake(1f).UpdateHealthBar();

meaning UpdateHealthBar MUST be at end

basically looking if there is a way to postpone UpdateHealthBar until chains are complete

[undone coral]

if you have autoplaying sequences, then yes

i'm pretty sure you can modify the sequence after you have called play anyway

[undone coral]

they're separate concerns

i mean look at the code

it shoudl be apparent that it's separate

[flint wraith]

hmmm doesnt seem that i change tween settings as vibrato, duration, strength and so on after i create it.

[flint wraith]

with late update it makes sense, i thought about that also but seems bit clumsier

[fair hound]

If I have a set of derivative classes I want a user to be able to choose between from within another class, how do I enable that ability? I want there to be able to be multiple instances of the same derivative type, but for each derivative type. My original idea was to have a private array of the base class, and a public array of each derivative type whose elements I feed into the base class array so I can iterate through them all from the one base array. Is there a better way to do this?

[violet valve]

Does anyone here know how to properly reference an asset and save changes to it through the editor? #↕️┃editor-extensions message

[jade grotto]

I have an IMeshModifier that offsets the verts of the Graphic but of course GraphicRaycaster is not aware. I temporarily move the Graphic's Rect where its verts get offset to just before GraphicRaycaster runs and back after, works well, but OnRectTransformsDimensionsChanged gets called and everything gets set dirty and performance is terrible

[runic girder]

public class BuildingDismantler : MonoBehaviour {
    [SerializeField] private Sprite _hummerSprite;
    
    private Entity Building => RaycastBuilding();
    private EntityCursor BuildingCursor => EntityCursor.instance;
    
    private bool _isDismantling;
    
    private void Update() {
        if (Input.GetKeyDown(KeyCode.Mouse0)) {
            TryDestroyBuilding();
        }
    }
    
    public void ToggleDismantleMode() {
        if (!_isDismantling) {
            _isDismantling = true;
            BuildingCursor.SetCursor(_hummerSprite);
            StartCoroutine(CheckBuildingContinuously());
        } else {
            _isDismantling = false;
            StopCoroutine(CheckBuildingContinuously());
            EntityCursor.instance.HideCursor();
        }
    }

    private void CheckBuilding() => BuildingCursor.SetColor(Building == null ? 0 : 1);
    private IEnumerator CheckBuildingContinuously() {
        while (true) {
            CheckBuilding();
        }
    }

    private void TryDestroyBuilding() {
        if (Building == null || !_isDismantling) return;
        Building.DestroyEntity();
        ToggleDismantleMode();
    }
    
    private Entity RaycastBuilding() {
        var hit = Physics2D.Raycast(Camera.main.ScreenToWorldPoint(Input.mousePosition), Vector2.zero);
        if (hit.collider.TryGetComponent<Entity>(out var building)) {
            return building;
        }
        Debug.LogWarning("No entity found.");
        return null;
    }
}```
Can this code be considered as a good one? I tried to follow all the principles.

[severe topaz]

@runic girder great, now switch over to NIS (new input system)
also Try without returning a bool makes me uneasy, usually you'd return it to make some kind of error sound if it fails

[runic girder]

NIS?

Gotta read about it

[severe topaz]

it's easier to switch early in the project, and it's definitely worth it

[runic girder]

This project is about to be released xd

[severe topaz]

well, if it's targeting only MKB (mouse + keyboard) then it's fine for now

[runic girder]

what's MKB?

[severe topaz]

@runic girder your ToggleDismantleMode has a check that's unreachable

trydestroy returns if !_isDismantling, so in toggle you don't ever get the first part

[runic girder]

I am actually going to use ToggleDismantleMode on button in editor, so it's fine, is it?

[severe topaz]

yeah, it's fine. i'm also not sure if coroutine stops if you do it like that instead of caching

i just tie everything to a bool

[runic girder]

Yeah, I've tested it, works fine

[severe topaz]

e.g. in coroutine { while(_isDismantling){ ... } cleanup }

[runic girder]

I actually tried to avoid it here. Wanted to use this coroutine only in specific moment

Otherwise this coroutine would run endlessly while checking If(_isDemantling)

[severe topaz]

wdym

it runs endlessly while checking if(true)

[runic girder]

yeah, but then it's being stopped when program toggles Dismantle Mode off

[runic girder]

Ohh, I actually just get it

[severe topaz]

i also don't see what's in CheckBuildingContinuously, does it return wait?

[runic girder]

It simply sets the building cursor's color

0 = red, 1 = green

yeah, and it converts bool to int

[severe topaz]

wait wait

while true, it sets color? and it doesn't crash?

[runic girder]

wellllll xddd

[severe topaz]

you didn't even test the code, did you? ahaha

[runic girder]

I actually didn't ahaha

[severe topaz]

it'll definitely crash ahaha

[runic girder]

easy to fix tho

[severe topaz]

in any case, if you don't need checkbuilding separately, i'd just move it right there, it's less confusing that way

private IEnumerator CheckBuildingContinuously()
{  _isDismantling = true;
    while (_isDismantling) { BuildingCursor.SetColor(Building == null ? 0 : 1);
      yield return CACHEDwait; }
}

[runic girder]

    public void SetColor(int index) {
        if(_cursor.color != _colors[index])
            _cursor.color = _colors[index];
    }

Fixed

[runic girder]

Yeah I already did it

but wait

what's yield return CACHEDwait?

[severe topaz]

if it's not null, you can cache it so that it doesn't generate garbage

for a building highlight you probably don't want it to be running at 120 fps, or even 50

[runic girder]

    private IEnumerator CheckBuilding() {
        while (_isDismantling) {
            BuildingCursor.SetColor(Building == null ? 0 : 1);
            yield return new WaitForSeconds(0.1f);
        }
    }

So you mean this, don't you?

[severe topaz]

cache it

static WaitForSeconds 100ms = new WaitForSeconds(0.1f);

then you can just return class.100ms; from anywhere

and it always sits in memory instead of dozens of copies that need to be cleaned up with GC (unless i'm totally misunderstanding how it works)

[runic girder]

static readonly WaitForSeconds CachedWait = new(0.1f);
private IEnumerator CheckBuilding() {
        while (_isDismantling) {
            BuildingCursor.SetColor(Building == null ? 0 : 1);
            yield return CachedWait;
        }
    }

this way?

[severe topaz]

yep

[runic girder]

btw, why should I make it static, Rider says I should do it, but why?

[severe topaz]

1: instances won't generate a new one
2: you can access it from other classes without a reference, instead of making a new one

[runic girder]

perfect, thanks

And thank you for such a good code review xd

[severe topaz]

yw, good luck with release

[runic girder]

thx!

[shadow vigil]

Does anyone know the math behind Transform.TransformPoint and Transform.TransformVector methods? I store positions in a struct, so I need something to perform the local to world transformation without the original transform.

[buoyant nacelle]

I thought about it yesterday. So, the event subscribes to a function inside an instance of a state. Then, i have a logic for movement inside of my state and have another class with animation that is called by the event

Right?

[violet anvil]

do we have any way to "intercept" any asset loading of some kind? e.g. when a model is loaded, get that model and add it to a massive buffer alongside many others, and prevent the "original" model load from actually being loaded into memory? my main goal here is to use said buffer for MultiDrawIndirect commands

[jolly token]

Look into Matrix4x4

[muted root]

Hi everybody, is it possible to use physics to grab parts of a mesh ? I have a character with a ponytail which has no bones inside. I was wondering if it was possible to create a script (or if an asset on the store already does that, I don't mind paying) which would allow me to move the ponytail with the mouse at runtime. Like grabbing the end of ponytail with the mouse, drag it around and it would follow the cursor. And if we grab it in the middle of the ponytail, the correct part would be dragged while the end of the ponytail would be affected by gravity and velocity.

The assets I found online only use either bone or the whole mesh

[jolly token]

Put some bone and skin it

[muted root]

Unfortunately, I'm not good enough to do that :x

[severe topaz]

there was a really nice asset on the store for that kind of thing

[jolly token]

It’s much harder if you try to do without bones

[muted root]

Hm, is it possible to add bones in unity ?

[muted root]

Oh? Would you have its name?

[lament salmon]

Another option is to use blend shapes, but for this you also need a 3D editing software. And bones will yield better results.

[lament salmon]

Though blend shapes might be easier to generate by code, than bones/bindposes

[severe topaz]

found it @muted root https://assetstore.unity.com/packages/tools/particles-effects/boing-kit-dynamic-bouncy-bones-grass-and-more-135594

[muted root]

I see. However blendshapes would feel less natural than bones or physics I guess.

[muted root]

OK thanks a lot, I'll check it!

[severe topaz]

besides that it's better to just make a bone though
just import to blender and do it from there
royalskies has a really nice and short tutorial on that
https://youtu.be/X590IZeyT1s

[muted root]

Yes but in my case, it's quite tricky. I'm using Daz3D assets, hair and cloth and editing them is really tricky, plus I'll have a lot of them so editing all of them would take a lot of time.

[lament salmon]

Have you considered using the Cloth component?

For ponytails and such.

[muted root]

yes, I've tested it and it does give a goof result but there is not grabbing feature from what I tested.

[lament salmon]

Ahh, I missed the "move with mouse" part

[muted root]

The ponytail does get proper physics and interact with chosen colliders.

[severe topaz]

either way yeah it's possible, but is it really a game-changing feature?

like, would you spend a month working on that?

[muted root]

Well, yes it would be quite needed since I'm aiming at an
"adult" game in the end where I'd like cloth to be removable with the mouse haha and hair to be pullable 😉

But when you say it's possible, I still don't see how it would work. All assets I found and even the one you linked are based on bones. So unless I modify hundreds of purchased asset to add bones, I still can't see a solutiion.

Which was why I was focusing on how to modify mesh partially, moving vertices individually but the math behind would be complex.

[severe topaz]

boing description says you just add component and it should work

no clue about it myself though

[mint spear]

https://youtu.be/gVgN5SU6BrA?t=2784

[severe topaz]

i'd just add bones since that's the most reasonable thing to do

[muted root]

From what I read "- The core logic for boing behaviors and reactors are simply modifying object transforms, so it should be compatible with all platforms Unity supports."

[muted root]

OK got it. I'll keep that in mind then when I have something more stable, it's not a critical feature as you said but would be a nice addition.

[severe topaz]

@muted root but on the other hand, boing is 50% off, although you can see hair breaking there

[shadow vigil]

Thanks! This is an amazing lecture.

[severe topaz]

@shadow vigil yeah, have fun binging on the rest of the videos ✨

[crystal galleon]

I have an issue with the duplication of a class and it's members. I've been looking at Deep Copy options and I'm having trouble coming up with a good one that meets my requirements. I'm duplicating a class that contains members of another namespace down-casted to it's in-namespace base class and I want it to be oblivious to what the derived classes are.

1. This means that probably Copy Constructor won't work, as that class' duplication method can't access the other classes in other namespaces. This is a limitation of assembly definitions, as it doesn't support cyclic dependencies.
2. I'm assuming that JSON Serialization creates a deep copy and it also is being a problem since one of the members of the duplicated class can be an UnityEngine.Object and is not Serializable. I do need to save that reference and It's currently declared as [SerializeReference] private Object owner (could add MonoBehaviour support in the future). On this option I tried to find the asset by name (all names are always unique), but doesn't really work as it doesn't let it run during serialization ({ get; set; })

[proper walrus]

at which point do I need to use LinkedPool? I find that ObjectPool is sufficient but is there any specific use case where I need to use it?
https://docs.unity3d.com/ScriptReference/Pool.LinkedPool_1.html

[midnight venture]

ObjectPool should more efficient, if you have more or less stable pool size and LinkedPool perform better, if amount of pooled object fluctuates.

[undone coral]

but why? what are you trying to do?

[undone coral]

you're overthinking this. you will never need this API. you will never need to animate the health bar from the outside, the health bar should animate itself

[jade grotto]

Curve UI around a cylinder, the curving works, the GraphicRaycasting doesn't

[undone coral]

can you share a screenshot?

[violet valve]

Beautiful!

[oblique crypt]

quick question so i have coded a state machine to control all of my movement walking running jumping also included flying and all the transition substates . now i want to add 3rd person shooter controls would it be better to expand this state machine or create a second set of scripts

[cobalt topaz]

Does anyone know how to blend two meshes into one. Like I wanna do an ocean with some very low vertex displacement with no subdivided ones. And then I have the waves around the player that should look high quality and very detailed and so on that plane is massively subdivided. How can I now blend the ocean with the player waves?

To get a smoother transition between high waves around the player and the low ocean waves

[flint sage]

Tesselation

Or 2 meshes

[cobalt topaz]

May you explain? xD

Like how can I do then tesselation?

And I need it as code.

[brisk pasture]

tesselation its a large topic might take a bit but its ideal for this case

[cobalt topaz]

Yeah, I know.

[brisk pasture]

though keep in mind you really do not need to combine or blend meshes

you can have 2 meshes look seamless

[cobalt topaz]

But I only see shaders for tesselation. But I need to script it.

[brisk pasture]

just as long as on the boundary the vertex positions and normals are the same

[cobalt topaz]

.

Yeah, I would need to script it.

[brisk pasture]

yeah what you are asking with the waves, doing it in shader is the ideal place to do it

[cobalt topaz]

But I need to make buoyancy. A boat should drive on it and a player should swim on it

And also like explosions and some other stuff....

[brisk pasture]

since simulating water is a lot of per vertex work, so vertex shader is ideal, or compute shader for just doing the work fast and getting some sort of output

[cobalt topaz]

But I need high waves

[brisk pasture]

it is a big topic, but you might be able search for ocean simulations and other stuff to get a grounding in it

[cobalt topaz]

A real stormy, dynamic water

I have done a lot of research

[brisk pasture]

yeah assuming you are wanting somehting like sea of theifs

[cobalt topaz]

And came to the conclusion that vertex displacement in scripts is ideal....

[cobalt topaz]

Similar

[brisk pasture]

yeah that is heavily vertex and compute shader driven

[cobalt topaz]

Yeah

But I need to dynamically recalculate the water height at x and z

Wait...

I could just do my general vertex displacement and then tesselation in a shader

Because I would get raw, low quality waves for physics but goodlooking ones as an actual output

[brisk pasture]

yeah could do the mix, generally physics wise it would be smaller

still might use a compute shader for that, since the resulting data from it you can use in code, but will run much much faster

though really depends on the resolution, if its a small enough amount of points it runs fine in C# can just start off with that

[cobalt topaz]

Yeah

But generally, how would I simulate those lots and lots of random waves...

I got Gerstner waves for now

[brisk pasture]

would have suggested gerstner or driving it with a texture

can get randomness changing some of its input paramaters at runtime or having multiple generations of them

[cobalt topaz]

Perlin Noise?

[brisk pasture]

could work, but even multiple generations of waves you scale down in height and multiply with previous ones can add a lot of variation

and let you have detail and multiple scales

[cobalt topaz]

Yeah

So, what I would do know, is vertex displacement for a radius of 11f, then tesselation from 10f to +11f.

[brisk pasture]

your asking the right questions, its just going to be fiddly to get the look you want

[cobalt topaz]

The topic is super interesting, in my view.

Like simulating nature... There's always some randomness.

[brisk pasture]

yeah, if you do try and multiply in noise, i would use perlin since you need things that do not jump too quikcly betwene values

[cobalt topaz]

Yeah, then simply Mathf Perlin Noise?

[brisk pasture]

depends on your code, and where the work is being done, but yeah should work, but you will have to play with what you pass for the x any y so the scale of the noise is correct for your usecase

[cobalt topaz]

X and Y, not Z?

[brisk pasture]

so those will have to be based on the location you are wanting to get the value for, but you may need to multiply them larger or smaller to loosen up the noise and get more macro scale detials or tighten it up

[cobalt topaz]

Yeah, I am gonna have a lot of public floats for tweaking xD

[brisk pasture]

but yeah since you mentioned needed different levels of detial you can fake it without tesselation

would just make a grid of meshes, where the ones closest to the player have a higher vertex density

the important part is just making sure the positions and normals of the edges match up with there neighbors

[cobalt topaz]

Yeah

But like that part of changing the vertex density is my Problem....

But the best way would be tesselation?

How would I do it then with a shader (having a circular shape)...?

[urban warren]

I have a question. I do very little multi-threading, and I was wondering how would be a good way to go about using multi-threading to create a mesh with a more advanced structure. In this case I mean a BMesh (specifically this lib https://github.com/eliemichel/BMeshUnity).
Like it requires adding items to a list and referencing things and stuff which multi-threading doesn't like.

Just kind of looking for a 'high level' overview of how one would approach it.

[sly grove]

adding things to lists and referencing things are not problems for multithreading in Unity.

Accessing objects that are managed by the engine from a background thread are.

So the specifically problematic part about using BMesh in a multithreaded environment would be that this method in particular needs to run on the main thread:
https://github.com/eliemichel/BMeshUnity/blob/3360cb6e2397dc0efb7137d6551ac60dd7dc6879/Library/BMeshUnity.cs#L48

[urban warren]

Yeah I know... I am kind of stretching the limits here of what is considered Unity and not just a normal C# question....

Doesn't adding things to lists in multiple threads cause problems in C# as two threads may try to put something in the same spot... or something...

(It really has been too long since I have done any multi-threading...)

[sly grove]

Yes adding things to the same list from multiple threads can be problematic

but that wasn't implied in the original question I don't think 🤔

[urban warren]

Was it not? I thought it was, sorry. It has a bmesh.AddVertex(..) method and a number of others that are used to construct the mesh.

Those of course, add the specified vertex/edge/face to a list on the mesh

[sly grove]

Yeah you'd have to stray away from those or use proper synchronization constructs around them.

The other approach is to preallocate the vertex lists and only Set vertices etc. in existing locations which you assign to different threads

[urban warren]

I see, that is kind of a large reason to use them.
All vertices are connected (reference) edges, which reference faces so generally you will need to cross reference things.

I guess generation can't really be multi-threaded too well.

Though modification can be.

[crystal parrot]

Any vue 3 user here?

[cobalt topaz]

How would you do a tesselation shader for URP?

(Based on camera distance)

[cobalt topaz]

Would this work, I don't think so cause It says that Surface Shaders aren't supported for URP? https://docs.unity3d.com/Manual/SL-SurfaceShaderTessellation.html

[surreal vessel]

am i crazy for not using a gamemanager?

idk just never liked the idea of it

could just be that its not useful in the game im making but it just seems to be a catalyst for bad code imo

[ashen shard]

It depends on what you're using instead of a gamemanager

[surreal vessel]

nothing really

everything controls itself

[ashen shard]

well

nothing crazy about that

[surreal vessel]

fair

i just see people mention it so often

and it makes me die inside a little

[ashen shard]

It is a really useful design pattern, but yes people tend to use it a bit too often.

For example in my game I have a tile/grid system that basically everything has to access, so it makes sense to use a singleton

[surreal vessel]

yeah true i suppose

fits some games more than other

[ashen shard]

What sort of game are you making?

[surreal vessel]

uhh

hard to describe

closest thing is probably a roguelike

[undone coral]

GameManager holds the parts of your game that are tightly coupled with each other. another example is Player, because a lot of people make games where input is tightly coupled with animation

[undone coral]

they are not supported yet

[surreal vessel]

i control that from the player itself

makes more sense to me than a third party manager to do it

[undone coral]

yeah, so if it makes sense to you why you have a Player class

that listens to input and then also drives animations

and does a bunch of other stuff

that is "coupled"

[surreal vessel]

nah its all separated components

[undone coral]

hmm

really depends on the game, it's an example

[surreal vessel]

i just have a reference to my animator in PlayerShooting and set stuff from ther

[undone coral]

the kind of game you're making decides whether or not something like animations are tightly coupled to input, not your class structure

does that make sense?

[surreal vessel]

yeah

[undone coral]

yeah that's all i'm really saying

if it's Spelunky, there is very tightly coupled player input, behavior and animation

most platformers look that way

compared to an RTS, where it is usually totally decoupled

you have to do the coupling, if it's right for your game, somewhere

some people like to suffer, and couple things using Dependency Injection

[surreal vessel]

oh hell no lol

[undone coral]

some people put the coupling in... a class. which in my opinion, is a good choice

of all the bad choices, it's a good one

a lot of people are making spelunky, then roll a state machine class, and it's hard

they have a state machine engine, and exactly one state machine (their player character in a platformer or FPS)

they have dependency injection, and exactly 1 implementation of every interface (service, or whatever)

[midnight venture]

Not sure if it’s actually that common things are tightly coupled in something like spelunky for example. I would guess it’s more like coupling through abstraction and for example all characters are ”built the same with different kind of components”

[undone coral]

you can only make that kind of retro/legacy/traditional 2d platformer input behavior by tightly coupling input and animation

how you choose to express that tight coupling is up to you

you can use Zenject to do that, and call it decoupled, and of course, it's actually coupled

you can also be in denial that there is some decoupled form of input behavior for that game, and if you try to make that, it won't "feel right"

there just isn't. all that retro platformer stick input movement behavior feels the way it does by harnessing the power of tight coupling

that's why i bring up RTSes, which have a huge variety of movement behaviors for units, and input is almost always as basic as clicking on a map

so of course that is decoupled

[midnight venture]

I’d say tight controls are no way related to tight coupling

[undone coral]

there aren't any RTSes with tightly coupled movement and input

[undone coral]

"tight" for a platformer... nobody would use the word tight to describe movement in an RTS

it's a very very retro thing

[midnight venture]

Spelunky like characters don’t care if their animation is driven by input or ai or whatever as long as the change is instant and there is no inherent delay in something like interfaces

[undone coral]

you can use Zenject to do that, and call it decoupled, and of course, it's actually tightly* coupled

you can certainly engineer those things however you want

kind of the core of @surreal vessel 's question

and i get it, you look at a GameManager and it has "all these lines of code" in it

many alternatives, for certain genres of games, are worse

[surreal vessel]

yeah and i suppose also depends on the scale of your game

my current project is prety ambitious so a gamemanager for one off things wont do

[undone coral]

another perspective is, you can find innovations when you go really hard one way or the other. like if you really want to make an innovative platformer, consider what player input would look like if it was highly decoupled

i'm not saying there is an audience for platformers with innovative controls - there problem is not - but you could do something cool there maybe

[undone coral]

well maybe gimme a 1 sentence logline, i know you said it was a roguelite (or roguelike)

[surreal vessel]

openworld, dungeon crawler, bullet hell, roguelike

[undone coral]

what i have seen is the levels are usually highly decoupled

because most of the innovation is in the encounters / enemies / terrain / items / etc.

and you build that up iteratively, and the interesting stuff sometimes happens when it's emergent behavior. like your player has the ability to turn land to water, and a water creature can now change its habitat, and this is emergent, it only happens serendipitously because the three systems - player inputs / "spells", terrains, and creatures - are decoupled

[surreal vessel]

yeah

things basically control themselves where possible

[undone coral]

it is not valuable to decouple the actually like, turning your joystick into movement

and animating the character in response the joystick

it just has to work a certain traditional way, that people expect

and there's pretty much no innovation there

that said, you can subvert expectations*

maybe it's a game where you take control of other creatures

[midnight venture]

I think you are just messing up terminology here and talking about something else than ”classes being dependant of specific classes” vs ”classes being dependant of abstraction”

[undone coral]

maybe it's a game where you can control things with your joystick, and there's something fun there, like the flow of water or the mood of a creature

you can also consider something tightly coupled for your world - maybe the world is generate from a set of storypoints you carefully design, maybe it's something like monument valley levels with npcs with specific dialogue

[undone coral]

it seems like we all know what i mean by tightly coupled. if you wanted to diagnose if something was "tightly coupled," for example, it's a method that is called in exactly one place only. paraphrased from real world:

void Update() {
 ...
 HandleInput();
 HandleAnimation();
 HandleSound();
}

or it's a code fragment that is called repeatedly in many places. that is something tightly coupled that is usually disguised as decoupled. some stuff i've seen in real codebases

enemy.GiveItem(item);
GameEvents.instance.OnItemGiven(enemy, item);

or

target.Health -= damage;
if (target.Health < 0) {
 target.DropAllInventory();
 target.Destroy();
 GameEvents.instance.OnTargetDies(target);
}

you can create a PlayerInput class and call it from Player... but the methods would only be called from one place

or you can have all that code in the same class

it's not about how many classes. coupling is the hard meaning of, would this thing only make sense together with this other thing?

most platformer input doesn't make sense by itself. it doesn't do anything by itself. for a long time, it was all of Input.GetAxis

[midnight venture]

Anyway, it’s not matter of coupling or using managers or whatnot if things are possible to do (like tight controls), but how simple or hard things get when you start expanding the system. No reason to do decoupling for sake of it, but it’s usually in the end simpler to design for example animation system which can be driven by anything instead of having one system for player, another for second player with a bit different skills and again different kind of animation systems for different kind of ai’s and so on

[undone coral]

yeah but you'd never need to expand something like a platformer's movement behavior. like either it is complete to spec, or it is not finished

there are basically no platformers that operate as services games, and none of those introduce "a new way to move around based on your joystick input" if that makes sense

for a 2d platformer, the design space of your level geometry is 😱 tightly coupled to your player input, because the movement of the player is

i'm not saying you have a piece of code that calls Input.GetAxis to determine how you, in your brain, lay out platforms

but that you have to account for how the player moves.

but how simple or hard things get when you start expanding the system
so my point is, you will just never do that. you have to do something like build your player input for a 2d platformer in 1 shot, and it has to be fully specified - maybe not done in an engineering sense, but a fixed design quantity - before you start building your level

and in my experience, this is why people building 2d platformers don't really develop skills ot build other games...

like shipping a 2d platformer is anti-experience

isn't that nuts?

[undone coral]

i don't know about animation either. the way megaman is animated, because it is so tightly coupled ti player input - it's like a special part of those games, those super traditional platformer inputs, how you read the animations and the specific pixels and whatever to know how the player behaves - none of that translates to an NPC

it isn't relevant

you don't control the npcs. so of course they can use a state machine for animation, which is much simpler

and i understand the megaman player has something that resembles a state machine, but the actual code for those games looks like a severe case of Updateloopese

where the update loop has very fine tuned interactions between which frame of which animation megaman is showing in response to which timers and which inputs

you're not wrong, it's just that these sorts of games, they don't ship with flexible animation systems

they could ship with them, and it would be wasted effort

and probably get in the way of achieving the thing those audiences want, which is almost uniformly megaman controls

can you tell i really hate the guts of 2d platformers

[midnight venture]

😂 Yeah, I get what you are saying and not denying this isn’t the case (especially for anything actually 8bit - you totally can’t look at that stuff through the lenses of modern game development), but funnily the roguelite part was one which threw me towards the ”go with the abstractions” route in this discussion. It’s quite common you actually want to expand lots of systems in games like that and have systems which allows you to develop crazy ideas easily

[undone coral]

that's true

[midnight venture]

So that one day when you want to give all the enemies skill to block like players block, or climb the ladders or give player ability to control enemies or something you don’t end up with spaghetti

[undone coral]

it's so hard to innovate in the levels & enemies without it being nicely abstracted

then again, every time i try to show people, "you're trying to make a CCG in addition to megaman, when you sign up to do an open world roguelike"

that gives the poor guy a heart attack

[undone coral]

you don’t end up with spaghetti
in my experience the best approaches are, can you make strategy card game text that describes the parts of your world you'd like to emergently combine together, and limit the design space of that text to something you can achieve in engineering

then you can avoid spaghetti - spaghetti is like, all the interesting card text you want to do, but you have to touch a bajillion things to achieve it

signing up to author XMage on top of your other game is a crazy move

XMage is nicely abstracted and has thousands of bugs

they are tractable to fix but nonetheless, if you wanted the Whole Design Space of Magic the Gathering, you're gonna ouch

we really gotta see the source of Rimworld

to know how far basic approaches to this stuff can go

[midnight venture]

Wellll, it’s not like physical card games don’t have bugs so to speak. Except in their case bugs are solved by house rules or developer blog posts or by additional rules enforced by tournament organizers etc

(Had to check, mtg rulebook is 278 pages long. I must say that is quite a lot for a game which tries to describe everything relevant inside a short card text)

[undone coral]

yeah hearthstone has a limited design space, like no change in priority, that poops away a lot of the complexity

paper prototyping a roguelite where input doens't matter as much is a Good Idea

you can definitely do this, as a creator, and make an innovative one

i am biased because i hate megaman controls, but what do i know, look at Shovel Knight

[severe topaz]

until you suddenly decide to try to add anti-gravity and/or vr
or to combine dash with attack that depends on weapon
there's always a game designer leaking creativity deciding to breaks previous mechanics in some way

[tender solar]

Hello, I am syncing my project to a new PC and am getting compiler errors relating to duplicate type definition in .Net and MSCorlib. here is one such error: ```Library\PackageCache\com.unity.services.core@1.0.1\Editor\Core\AsyncOperation\IAsyncOperationAwaiter.cs(35,47): error CS0433: The type 'ICriticalNotifyCompletion' exists in both 'AsyncBridge.Net35, Version=0.2.3333.0, Culture=neutral, PublicKeyToken=b3b1c0202c0d6a87' and 'mscorlib, Version=4.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089'

Any idea what might be causing this and how to fix? Presumably I can't edit MSCorlib.

It's also working just fine on the original PC the project was made for.

[severe topaz]

duplicate files, possibly in different folders

[tender solar]

So it's definitely something within the Unity project?

[severe topaz]

wdym unity, it's a C# error

it's trying to rebuild it but found 2 types with same name

[tender solar]

Hmm. Updating Visual studio and re-syncing the project from github seems to have fixed it.

Though now the .sln says I need to update Unity for Visual Studio to use any code tools.

[spice wave]

Is there some callback function for when an animation transition has ended?

[sly grove]

There's https://docs.unity3d.com/ScriptReference/StateMachineBehaviour.html
And there are animation events

[reef lance]

I am making a server plugin for the game Rust and I want to display some fancy GUI, I know that rust uses Unity's UI system and I want to send a command to the client to draw something, this is done via RPC. Now the problem is that I dont know what data to send, so I started reverse engineering and now im stuck at some method that is loading some unity asset in which I pressume there is a value which was set in the inspector because in the decompiler the field is not initialized

So the tldr is where is that inspector value stored and can I somehow see it?
I tried almost all unity resource extractors/resource viewers none of them give the thing I want

Any help appreciated!

[cobalt topaz]

What can I do instead? i really just want high quality waves (high subdivision) in a radius and outer waves around that aren't subdivided.

And the radius position would be the same as the player's.

[cobalt topaz]

Like here: https://forum.unity.com/threads/subdividing-a-mesh-through-shaders.526323/

[undone coral]

what are your target platforms?

[cobalt topaz]

Windows

[undone coral]

how do you feel about using hdrp

[cobalt topaz]

I don't mind. I can use it too.

I just want to get it working.

[cobalt topaz]

Really, it doesn't matter.

[undone coral]

well hdrp has tessellation support

it is much closer to the Unreal / high fidelity gaming pipeline

[cobalt topaz]

But how can I do it then?

And here? https://docs.unity3d.com/Manual/SL-SurfaceShaderTessellation.html

It says that HDR is unsupported.

[undone coral]

what is your objective?

what are you trying to make?

https://youtu.be/ycJ434Lh21w?t=583 here is an example of tessellation in hdrp

i'm worried you're talking about surface shaders

[cobalt topaz]

.

I want that.

Highly subdivided waves in a radius (with a shader), and out of that radius no subdivided waves.

[undone coral]

when you say waves, you mean you want to make an ocean?

[cobalt topaz]

I did that.

[undone coral]

like the purpose is to make a water effect?

[cobalt topaz]

Let's say a plane

[undone coral]

are you doing this using shader graph?

[cobalt topaz]

No, code

[undone coral]

so

[cobalt topaz]

Because I need it so that objects can swim

Vertex Displacement

[undone coral]

i think you can use built in render pipeline then, and proceed along

[cobalt topaz]

Okay, and then use Surface Shaders?

Imagine Sea of Thieves water....

[undone coral]

there's a reason surface shaders were eliminated from hdrp

[cobalt topaz]

ok

[undone coral]

and why they haven't really bothered with reintroducing them in URP

[cobalt topaz]

ok

[undone coral]

i can't tell if you're interested why

it sounds like you want to write shader code

which is the problem

[cobalt topaz]

No, it is not

[undone coral]

you don't want to write shaders or you do?

it depends on the goals

[cobalt topaz]

I don't need to

https://docs.unity3d.com/Manual/SL-SurfaceShaderTessellation.html

Just copy and paste

[undone coral]

if your goal is to create attractive looking water

[cobalt topaz]

I think you don't understand me.

The water shader is working and looks good.

I just need tesselation for extra smoothness

[undone coral]

hdrp has a water surface you can experiment with and extend

its behavior is similar to what you find shipping in many games, and is meant to be a clone of whatever unreal has

[cobalt topaz]

But I need high waves and low ones-

I need to dynamically generate those

Low waves around islands and then higher ones based on distance

[undone coral]

there might be assets that do this too

[cobalt topaz]

But when I use the buit-in, I don't have shader graphs right?

So, my written water lit shader won't work

[undone coral]

i don't understand, did you write a coded shader, or did you make a shader graph shader

i think you'll figure it out sorry i can't be of more help

[cobalt topaz]

shader graph

[twilit sable]

Instantiate(pmag, parent: Magparent.transform, position: new Vector3(0.0657f, 1.3951f, 1.0043f), rotation: rotation2);
How to make it instantiate in local certain coordinates (inside magparent)? Now it spawns in world position (изменено)

[sly grove]

As I told you yesterday - you can't. You have to give it a world position.
So you could convert your local position to the world position you want. E.g.:
Magparent.transform.TransformPoint(new Vector3(0.0657f, 1.3951f, 1.0043f))

weird to have a hardcoded offset like that though

recommend just adding an empty object as a child of Magparent at the correct location

and use that position directly

e.g. theEmpty.transform.position

[twilit sable]

Ok, thank you, it is hard to undestand for me

[jolly token]

I'm not sure how is this advanced topic, ...for example just rotate offset vector and add your player's position?

[past silo]

Oh, sorry, its advanced for me :D. I have never used the new input system with controller, only for simple Keyboard movement. But ill move my question.

[junior nest]

i am codegenerating a class and then trying to create a scriptable object from that class. However i need to wait until the assetdatabase refresh finished until i can actually create the scriptable object. I thought executing this in [DidReloadScripts] is the way to go but ScriptableObject.CreateInstance() does not create an instance because it cannot find the Type yet.

[DidReloadScripts]
        public static void CreateScriptableObjectAsset()
        {
            var folderPath = "Assets/Scripts/UtilityAI/CodeGen/AutoGeneratedRefs/";
            if (newCodeGenTypeNames.Count == 0) return;
            foreach (var newTypeName in newCodeGenTypeNames)
            {
                if (!string.IsNullOrEmpty(newTypeName))
                {
                    Debug.Log("Trying to create ScrObj for typeName: " + newTypeName);
                    var newScrObj = ScriptableObject.CreateInstance(newTypeName);
                    AssetDatabase.CreateAsset(newScrObj,
                        folderPath + newTypeName + ".asset");
                    AssetDatabase.SaveAssets();
                }else
                {
                    Debug.Log("newTypeName is null ");
                }
            }
            newCodeGenTypeNames.Clear();
        }```

anyone knows how i guarantee that all newly created types can be discovered?

I know the code generally works because when i restart the editor every asset is generated correctly. Its just a timing issue

[compact ingot]

maybe with an asset postprocessor combined with domain reload detection? https://docs.unity3d.com/ScriptReference/AssetPostprocessor.html

[junior nest]

very useful! i didnt know about them. i have the same issue however...

UnityEngine.ScriptableObject:CreateInstance (string)
UtilityAI.ReferenceGeneration/MyAllPostprocessor:OnPostprocessAllAssets (string[],string[],string[],string[],bool) (at Assets/Scripts/UtilityAI/Core/ReferenceGeneration.cs:139)```

this error pops up once but then with PostProcessor it now runs again and creates the object

so its still not waiting just right

[compact ingot]

maybe this helps: https://docs.unity3d.com/2021.3/Documentation/ScriptReference/AssetDatabase-importPackageCompleted.html

its typically used to display those "hey you installed package Foo, wanna look at our website" windows

[junior nest]

in my case it wont help. i need to create scriptable object instances that hold a reference to a type that has a implemented a certain interface. those types with the interface are outside of packages and my AI package is trying to reference them

i know sounds stupid but there is an idea here 😄

[compact ingot]

i don't pretend to understand

but if its just a question of waiting a bit, why not use an editor coroutine?

[junior nest]

but what to wait for? i kinda dont want to use wait 5 sec. (especially since it might be that i create quite a few such classes in some rare cases and it could take longer)

[compact ingot]

idk, no idea why the type isn't loaded after the domain reload

maybe this detailed illustration of domain reload contains a hint: https://docs.unity3d.com/Manual/ConfigurableEnterPlayModeDetails.html

maybe this event? https://docs.unity3d.com/ScriptReference/AssemblyReloadEvents-afterAssemblyReload.html

or poll this property https://docs.unity3d.com/ScriptReference/EditorApplication-isCompiling.html

[junior nest]

its crazy. it does not compile but it doesnt find the script

[compact ingot]

is it referenced by, or part of, the assembly you are operating in?

[junior nest]

yes im creating that asset inside my AI assembly where that script is running too. as i said everything is actually created correctly. it just throws this error once before creating it

@compact ingot thanks a bunch for your help! i learned alot

need my bed now 🙂 ill find a way tomorrow

[undone coral]

so is there any way to profile without a development build yet

i think there is an experimental runtime profiling package right?

[warm forge]

the profiler?

https://docs.unity3d.com/Manual/Profiler.html

[manic flare]

deep profiling

[broken cave]

hello, has anyone tried synchronizing the selection within UnityEditor.Experimental.GraphView with Selection.objects so you can edit them in the property editor (in my case, nodes are backed by ScriptableObjects)? selecting elements in the GraphView correlated to Selection.objects when Undo.undoRedoPerformed is triggered doesn't seem to help. I see the selection showing up in my undo history, but the GraphView doesn't update with the selection, even though I called AddToSelection with the right elements, which exist in the GraphView.

[junior nest]

i am doing it with Selection.SetActiveObjectWithContext(scriptableObjectData, scriptableObjectData);

[wet burrow]

More advanced C# than advanced Unity, but it's worth asking here.

public struct CategoryTag
{
    [JsonProperty("id")] public int id;
    [JsonProperty("name")] public string name;

    public static bool operator ==(CategoryTag a, CategoryTag b)
    {
        return a.id == b.id && a.name == b.name;
    }
    public static bool operator !=(CategoryTag a, CategoryTag b)
    {
        return (a == b) == false;
    }
}

I want to use the equals operator with this struct, but it doesn't recognise them so I had to make my own. This caused a green squiggly line over the struct name and these errors:
'CategoryTag' defines operator == or operator != but does not override Object.Equals(object o)
'CategoryTag' defines operator == or operator != but does not override Object.GetHashCode()
Is this a problem? Can I override the GetHashCode() function and then not need to make the custom operators? What's the best way to implement something like this?

[pliant crest]

just override equals (as well)

[wet burrow]

But in that case what would I override it with? I know for GetHashCode() I'll need to make a custom function to generate one from the variables, I presume I'd have to make new code for the Equals() override. Simply checking 'equals' for each variable within the struct sounds so easy that there has to be some kind of catch.

[pliant crest]

since its a struct

wait hold up

why are you overriding the == anyway?

[flint wraith]

well if you want to check if they are IDENTICAL CategoryTag , like same values then yes thats what you do

but it would return true also if object is not the same

well not object, since its struct

[plush hare]

    {

        Transform[] _Transforms = _Transform.GetComponentsInChildren<Transform>();
        for (int i = 0; i < _Transforms.Length; i++)
        {

            T[] _Components = _Transforms[i].GetComponentsInChildren<T>();

            for (int j = 0; j < _Components.Length; j++)
            {

                _Components.enabled = true;

            }

        }

    }``` How do I do this? Component doesn't contain enabled

[pliant crest]

the override is basically pointless

so i don't see any reason for them to do it

but if you really want to

you can just create a unique hash from the fields

that you need to compare

(which in this case is all of them which is what it already does)

[flint wraith]

yeah was just thinking, struct doesnt have instance

[potent wraith]

Anyone have experience in changing a mesh filter of a player object. I'm trying to do a prop hunt effect. Here is the current code, using a raycast to get the object which is working. But switching the meshfilter of the player is not working It should just be be a Chair for the mesh filter.
Code:

if (Physics.Raycast(transform.position, transform.forward, out var hit, Mathf.Infinity, mask))
        {
            var obj = hit.collider.gameObject;
            Debug.DrawRay(transform.position, transform.TransformDirection(Vector3.forward) * hit.distance, Color.yellow);
            Debug.Log($"looking at {obj.name}", this);
            Mesh mesh = obj.GetComponent<MeshFilter>().sharedMesh;
            Mesh mesh2 = Instantiate(mesh);
            player.GetComponent<MeshFilter>().sharedMesh = mesh2;

This is what the objects mesh filter changes into(it is supposed to be Chair)

[potent wraith]

Figured it out, Its because the mesh I was transforming into was marked as static 🤦‍♂️

[quiet bolt]

Lol

[sharp lake]

https://docs.unity3d.com/ScriptReference/Behaviour.html

[plush hare]

Tried that. Many components apparently don't inherit from Behaviour and they break when you try to treat them like a behaviour

Character controller, mesh renderer are a few I tried. Camera seems to be one of the only things that actually worked with that.

[sharp lake]

by treat u mean inherit them from Behaviour?

[plush hare]

I would get the component using GetComponent<T>() and then cast that to a behaviour

[sharp lake]

u probably need to inherit them from behaviour, might be annoying for overriding all default components u need, but im not sure if thats how u supposed to do this

[plush hare]

I gave up. Instead of disabling them, I just deleted them using the script I showed

[jolly token]

Override Equals, check other is CategoryTag otherTag, and if so return this == otherTag. If not return false. The point of Equals override is that it should return consistent result as == operator.

For HashCode simple way is return field1.GetHashCode() ^ field2.GetHashCode() ^ …;. You can choose better way to combine hash to reduce hash collision, but it works as long as same object returns same hash.

[radiant junco]

Hi all. While creating a mesh by script, do I need submeshes to setup different materials/uv?

[regal glade]

is it a bad idea to have lots of rendertextures around or does it not really matter until you hit some critical limit? I read about 200+ for certain mobile devices would cause crashes. I'm running the objects through a computeshader every update, and I'm just not sure I want to copy the pixels from a rendertexture to texture2Ds 50 times per frame instead of just setting RenderTextures as _MainTex in the shader.

[sage radish]

I think it's just the memory usage that matters, which copying them to Texture2D won't help with, because you're not compressing them by doing that and they still need to be in GPU memory to be accessible to shaders.

[regal glade]

hmm alright, thanks.

[frozen badger]

Hey, I've got an interface that need an ID attached to them in order to function correctly, however interfaces can't hold fields. Is there a best-practice solution to get something like a field into an interface? Currently I've got a property that I implement in the classes using the interface, which I don't want to rely on everyone doing.

[sly grove]

Properties in interfaces is fine

[frozen badger]

Yeah but I can't put the backing field in the interface

[sly grove]

Correct

Use an abstract class if you want a field

You can even use both 😉

[frozen badger]

Yeah but the implementing classes are too different to reasonably inherit from the same object

So basically I have
public int deckObjectID { get; set; }
in the interface, and then every class has to implement the getter and setter on their own

Which is a lot of duplicate code that I'd love to have in the interface, especially considering they all get their ID from the same supplier

[sharp lake]

but do u need to "implement" it, basically u just need to copy exactly same line of code, right? since get; set; will be already an implementation by itself

its not that bad

[sly grove]

You should use a separate component to hold the ID for this system then

Unity thrives with composition rather than inheritance

[frozen badger]

These are not components on gameobjects, unfortunately

Basically they all need this, and also in their constructor they need to set the backing field correctly

[sharp lake]

i mean do u need that variable?

what if u use deckObjectID instead of it

u can use getter/setter as property itself

[sly grove]

What are they then

[frozen badger]

Huh, I didn't know I could just re-declare the property with default get set and then that would be fine

[frozen badger]

Some of them are normal code-only objects, some of them are scriptableobjects

[frozen badger]

Then I at least don't need the backing field, which is slightly better

[sharp lake]

tbh thats a little weird to have one specific property in such different classes, no idea how u gonna use that property

[frozen badger]

Basically I have a custom collection I called a "Deck"

[sharp lake]

ye its basically 1 line of code per class, isnt that bad

[frozen badger]

I'm still working on it, but basically it's a set of dictionaries controlling the current state of that object

I want to be able to reliably control whether or not items in my deck are "available" to be rolled, locked, which means they have been rolled as an option but not claimed yet, or claimed, which means it's active in the game at the moment

I want to be able to o(1) get an item from the dictionary so every object needs to supply an ID, basically

[undone coral]

huh?

i wouldn't overthink it. how many cards are in your deck

there is a reason KeyedCollection with multiple keys doesn't exist

[frozen badger]

It depends on the deck. I named it "Deck" but not because the contents are exactly "cards"

[undone coral]

it is possible to author 100% correctly in typescript, but not 100% correctly in c#

well

how many

[frozen badger]

For the Item-deck it could be like, 100?

[undone coral]

okay

[frozen badger]

For the hero-deck probably like 15

[undone coral]

i don't think this optimization will yield much

100 items is very small

i don't understand what this is supposed to store

[frozen badger]

It's a custom collection so it stores a bunch of things

But with pre-coded functionality for keeping track of what has been "pulled" already

[undone coral]

okay

[frozen badger]

Like if all of this was just in a list I'd need to manage that separately, probably for each type

[undone coral]

hmm

you keep adding more and more stuff that makes it sound like you've veered off very far down Bad Lane

[frozen badger]

I may have

[undone coral]

do you want to get out of the Bad Neighborhood

[frozen badger]

like, when a hero is rolled I don't want to be able to roll him again for the rest of that game

I'm hoping I can black box the bad neighborhood

[undone coral]

or do you want to become a boss of crimes against programming

[frozen badger]

like build a giant wall around it

[undone coral]

lol

[frozen badger]

and then I don't have to look into it so much

[undone coral]

but that is so exceedingly simple, that there is definitely something i don't understand yet

is this like hearthstone's roguelike format?

[frozen badger]

Like, this is not a card game, but I want the randomness to work the same way as a card game does

[undone coral]

like whatever it's called

Card Game Roguelike Where you Go Up A Tower and there's Poison

[frozen badger]

like let's say you pull 3 cards, you pick one and the other 2 go back in the deck

and when a card is pulled, it can't be pulled again

[undone coral]

yeah

that makes perfect sense to me

[frozen badger]

but this is not a "card" game

[undone coral]

well what kind of game is it

[frozen badger]

and I want many different things to follow that same randomness pattern

[undone coral]

i make a lot of games

you can tell me, i will understand

"i hear you"

[frozen badger]

it's closer to an RPG, but the games are shorter

you win or lose in an hour or two

[undone coral]

okay

you mean it's classic final fantasy style combat, and you're pulling encounters from this deck?

is that the idea?

[frozen badger]

and the game world is pre-set with items and monsters in these decks that run out over the course of the game

[undone coral]

yeah

okay i hear you

[frozen badger]

It's not classic ff combat, but I am pulling encounters from the deck

[undone coral]

that makes sense okay

[frozen badger]

and items, and which characters show up, etc

[undone coral]

so you want to have a Discover mechanic too right?

[frozen badger]

Yeah

[undone coral]

okay great

[frozen badger]

like "Show me 3 healing items and let me pick one of them"

[undone coral]

"put the other two back"

makes sense to me

[frozen badger]

exactly

And since I have many things I want to behave like this

[undone coral]

yeah

i get you

[frozen badger]

I implemented my own collection for it

[undone coral]

so where does the initialized list of monsters and items for this playthrough live?

in a List?

[frozen badger]

which takes care of all of this generic pulling, locking, "shuffling" etc

Well, right now it is in a dictionary

because I wanted to be able to remove a specific instance of a "card" so to speak

[undone coral]

yeah

that has made things way harder

a deck is a list not a dictionary

which is okay

you are reimplementing a list on top of a dictionary right now

[frozen badger]

I am not overly worried about looping over the keys etc

randomly getting a key index

However, each item needs to know its own key

[undone coral]

why?

[frozen badger]

well, otherwise how can I know which item to move around

[undone coral]

basic equality checks

[frozen badger]

It just.. seems wrong to directly compare a complex object

[undone coral]

why? the references will be the same

it won't be complex at all

[frozen badger]

You're saying I loop over my list to find the current object and then move that, right?

[undone coral]

sure, or whatever you need to do

[frozen badger]

like, myList.Remove(myChosenItem);

which is fair

[undone coral]

yep, that will Just Work

[frozen badger]

I'll still need my subclasses for this deck though

as they have different types of discovery

[undone coral]

create different methods

[frozen badger]

but it does get rid of the ID part

[undone coral]

don't make multiple classes

you don't need them

just make differently named methods for discover

i'm writing the example hold on

[frozen badger]

This is a Deck<T> at this point

[undone coral]

it might not be necessary. you don't actually need to know anything about the T

[frozen badger]

so Items can be looked up by their complete name, their item tags

[undone coral]

i mean of course it's a collection, you can have it be T

[undone coral]

that's find, but then have a class MonsterDeck : Deck<Monster>

don't do new Deck<Monster>

but you won't really need to do that. you can put those methods anywhere. hold on a sec let me show you an example

[frozen badger]

For stuff that requires no extra functionality I just use Deck<T>, and then for Items for example I use ItemDeck<T>

[undone coral]

no

just do MonsterDeck and ItemDeck

like i'm showing

that implementing the concrete type

generics are not for extending behavior, they are for collections and that is all

if it's an ItemDeck, it's an ItemDeck, not an ItemDeck<T>

it's not an ItemDeck<Potion>

it's a PotionDeck

you can have an ItemDeck<T> as long as it is abstract

but i think that's a bad idea

just put the methods

that have to deal with these things somewhere else. not in the collection

it will be too confusing to put it there

[frozen badger]

But, would you have these inherit from Deck?

[undone coral]

sure

but you don't need this

[frozen badger]

public class ItemDeck<T> : Deck<T> where T : GameItem

[undone coral]

lemme show you an example

no

don't do that

[frozen badger]

As it is it's only an ItemDeck<T> so it can extend Deck<T>

[undone coral]

here's an implementation of a deck

sealed class Deck<T> : IReadOnlyList<T> {
 private List<T> entities = new();

 public Deck(List<T> startingEntities) {
  entities.AddRange(entities);
 }

 public void Shuffle() {
  // shuffle the entities array
 }

 public List<T> Draw(int quantity) { ... }

 public void ShuffleInto(List<T> cards) { ... }
}

// elsewhere!!! the smarts of this is interacting with the UI
Deck deck;
public async UniTask<T> Discover<T>(int count) {
 var drawn = deck.Draw(count);
 var chosen = await GameController.instance.userInterface.ShowAndChooseFrom(drawn);
 // now not drawn
 drawn.Remove(chosen);
 var notDrawn = drawn;
 deck.ShuffleInto(notDrawn);
 return chosen;
}

[frozen badger]

Because all the rolling, locking etc of items is done one time in Deck, but sometimes I need access to stuff in GameItem etc

IReadOnlyList?

[undone coral]

you don't have to use it if you don't want

it won't matter

because all deck needs to be aware of is whatever is in common with monsters and items. which should be the base class of items and monsters

[frozen badger]

Right, but the reason I am using T here is because they have no shared base class

except for like, Object

[undone coral]

then it's an object!

[frozen badger]

they are fairly far apart

[undone coral]

well

okay this is fine to use for T, but do not make any other specializations

[frozen badger]

Right, but then in a subclass

like, ItemDeck

I want to be able to access item-specific discovery

like discovery based on effect

or rolling an item based on rarity

[undone coral]

the problem is that it causes all of your methods that operate on this, in interesting ways, to become typed in a way that is useless to you

[frozen badger]

what do you mean?

[undone coral]

it's hard to explain because you have no experience in this and i have a lot

[undone coral]

just Write a Method to do that

like don't overthink this

of course you can use generics

but don't put this functionality of rolling an item based on rarity

into the Deck class, or any subclasses of it

write the method Anywhere Else

does that make sense?

[frozen badger]

Right, but that is the thing I wanted the deck to do..

[undone coral]

because you're going to have only one instance of an ItemDeck

[frozen badger]

sure

[undone coral]

yes

but think... where is that one instance of ItemDeck going to live?

[frozen badger]

In some manager class

[undone coral]

yes

so write the method in the manager class

[frozen badger]

but then in a sense, this deck is that manager class

the item deck is the manager for items

it's the "item manager", right?

[undone coral]

no

okay

look at my example

look at Discover

do you see the problem with it right now? the problem was introduced by adding <T>

[frozen badger]

Yeah, your deck is more of a pure container

[undone coral]

no

Deck deck;
public async UniTask<Entity> Discover(int count) {
 var drawn = deck.Draw(count);
 var chosen = await GameController.instance.userInterface.ShowAndChooseFrom(drawn);
 // now not drawn
 drawn.Remove(chosen);
 var notDrawn = drawn;
 deck.ShuffleInto(notDrawn);
 return chosen;
}

versus

Deck deck;
public async UniTask<T> Discover<T>(int count) {
 var drawn = deck.Draw(count);
 var chosen = await GameController.instance.userInterface.ShowAndChooseFrom(drawn);
 // now not drawn
 drawn.Remove(chosen);
 var notDrawn = drawn;
 deck.ShuffleInto(notDrawn);
 return chosen;
}

what becomes way more complicated because of <T>?

this should illuminate why you need an Entity - you can have something else, but the alternative is way too nuanced to discuss right now

[frozen badger]

Ah, but the actual UI part of this wouldn't be in the deck

[undone coral]

i know!

it's not

this is "somewhere else"

it doens't matter where it is

but that's what's valuable

and do you see

like writing the collection manipulation logic takes milliseconds

writing the UI is painful

[frozen badger]

Basically right, I need the base class Deck to work on all types

[undone coral]

and because you introduced T, ShowAndChooseFrom<T>(List<T> listOfAny) becomes way more complicated than ShowAndChooseFrom(List<Entity> cards)

you need a base class for your items and monsters

[frozen badger]

sure

So you're saying, in Deck, treat everything as Object

[undone coral]

and 99% of the time, your UI will not care if it's an item or a monster. it will care what it should show where

[frozen badger]

and then in a subclass

treat everything as the class I need it to be

[undone coral]

no, in deck, treat everything as an Entity

and make a sealed class Deck<T> where T : Entity if you want to have specialized collections

a lot of people make this mistake

they think about the collections being specialized and adding functionality

you'll notice in the CLR there is no ListForItemsThatAreCards<T> : List<T>

that's just not what generics are for

it feels like it's what they are for

they're not for grouping a bunch of methods together. do that somewhere else

put it in your game manager

[frozen badger]

Hm

[undone coral]

it makes sense for the deck to have a shuffle method

but also, it does not. trust me. in my game, i don't have Shuffle in Deck

[frozen badger]

It also makes sense for the deck to have a draw random method right

[undone coral]

no

[frozen badger]

like what would you have in the "Deck" at all?

[undone coral]

for something like hearthstone, just additional data about the deck

[frozen badger]

if I take away the draw, the discovery, the specialized functionality, the search

right but this is not a card game

[undone coral]

right

the reason those things should live in game manager

[frozen badger]

the "Deck" is specifically there to be a collection of functionality

[undone coral]

is that people will do stuff, like use an item, that changes the rules

something like Shuffle is tightly coupled to the rest of your game

[frozen badger]

regarding randomness for objects as if it was a card

[undone coral]

for example, you might have an item that skips shuffling, or an item that lets you keep all three items you discover, etc.

[frozen badger]

Right but as I said this is not a card game

this is only randomness as cards work

as opposed to say how dice work

[undone coral]

yes, and i am saying, actually, all these games have tightly coupled functionality

[frozen badger]

if you roll 1 on a dice, you can roll 1 on your second dice

[undone coral]

you will wind up with the deck having to peek at the item - this is a problem you ALREADY have

which it shouldn't do

[frozen badger]

but if you pull the ace of spades, you can no longer pull the ace of spades, right?

[undone coral]

that's just specific to your game

i'm saying the kind of game you're making

you will wind up having items and monsters or other stuff

that will make 1 change to the rules

that will require an invasive change to your Deck class

[frozen badger]

Right but none of those rules can be "prevent shuffling"

because the deck isn't there to "be shuffled"

it's not an actual deck of cards that are say, pulled each turn

[undone coral]

maybe it' snot shuffling but it can be many other things

the key problem you came here with is using a dictionary instead of a list

[frozen badger]

Sure

[undone coral]

you actually have everything else under control. don't declare a Deck class at all really

[frozen badger]

and it is probably better to do a list

[undone coral]

just use a List

and any specialization, write a method somewhere else. in GameManager

[frozen badger]

But I need a bunch of different objects to then control the randomness

[undone coral]

here's my deck - https://github.com/hiddenswitch/Spellsource/blob/master/spellsource-game/src/main/java/net/demilich/metastone/game/decks/GameDeck.java

[frozen badger]

like, locking drawn cards

so they are not drawn again

[undone coral]

you just remove items from a list

[frozen badger]

right but then the cards that are not chosen have to be put back

[undone coral]

i mean, we can talk about how to implement that with a list

it's pretty easy

well

i showed you a Discover implementation

[frozen badger]

Sure

[undone coral]

here it is with a plain list

[frozen badger]

Right, but in terms of our "decks" I have very few of the variables in yours

the deck has no descriptor, no name, is attached to no player or hero

it also has no ID

[undone coral]

public async UniTask<Entity> Discover(IList<Entity> deck, int count) {
 var drawn = Draw(deck, count);
 var chosen = await GameController.instance.userInterface.ShowAndChooseFrom(drawn);
 // now not drawn
 drawn.Remove(chosen);
 var notDrawn = drawn;
 ShuffleInto(deck, notDrawn);
 return chosen;
}

then you don't need a Deck

here's drawing a card: https://github.com/hiddenswitch/Spellsource/blob/9bca0a14ac64dbbbed3e9e919b0a5df30030b564/spellsource-game/src/main/java/net/demilich/metastone/game/logic/GameLogic.java#L2052

[frozen badger]

Right, but having all of that stuff in a game manager

controlling how to draw items, events, heroes, monsters, etc

becomes a very big game manager

[undone coral]

i don't know what to tell you

[frozen badger]

and it's not even the main part the game manager would do, right?

[undone coral]

these games are complex

[frozen badger]

It seems like the name "Deck" confuses it

it's really a system for controlling the randomness

not at all related to actual cards

[undone coral]

you can make something worse, that takes 5,000 lines and distributes it int hte wrong place, among 20 files, that have to invasively know about everything anyway

and then you wind up with 15,000 lines

[frozen badger]

merely following the same pattern of randomness as a deck of cards would

[undone coral]

anyway this is my perspective, i have a lot of experience in this. ultiamtely you are leaning on

the psychological ickiness of looking at a file from a mature, complete game

[frozen badger]

and I wanted to separate this functionality out into its own class specifically controlling that

[undone coral]

saying it's "too big"

like you know in your heart of hearts i'm right

[frozen badger]

Well, it's a question of "where do I put it" right

it definitely could go both places

[undone coral]

your game rules should go in one file.

they are tightly coupled

[frozen badger]

yeah but this is not a part of my games rules

as my game is not a card game

[undone coral]

if you have any spell like functionality in your game, i am telling you now, it is very invasive

[frozen badger]

like imagine that I am making world of warcraft right

[undone coral]

it IS part of your game rules

[frozen badger]

and instead of a loot table, I would use card-like randomness

I would not put this item rolling functionality into the game manager, right?

[undone coral]

i think it's the best place to put it

[frozen badger]

Even though it could be there, I would probably separate out everything related to rolling loot into its own file

[undone coral]

for the same reason that basically every collection type omits functionality like this

[frozen badger]

Well, I could indeed ditch the T

and just use "object" in the base class

[undone coral]

okay well

i think there's a lot for you to digest here

[frozen badger]

so it would be Deck that just constantly refers to "Object"

Yeah, thanks for the input

[undone coral]

you don't necessarily need Entity, you need Renderable

[frozen badger]

renderable?

[undone coral]

and you also need HasId

yeah because you're going to have a method in your UI

that takes a thing and tries to show it on screen

sometimes it's an item and sometimes it's amonster

[frozen badger]

Ah, no, these do not go on screen as-is

[undone coral]

and all the hard part of making a discover ui is showing the three rectangles and setting up the event handlers and doing the animation

hmm

[frozen badger]

as they are not actually cards

[undone coral]

well i have been asking you

what kind of game this is

[frozen badger]

they go backstage

[undone coral]

it will help me give you more specific examples

[frozen badger]

Sure

[undone coral]

it's still got Spells

as soon as your game has Spells in it, conceptually speaking

[frozen badger]

Basically I have UI elements that are fed backstage objects

[undone coral]

it will look like this

[frozen badger]

if you imagine it was a card game and was hearth stone, I would have the "Card" and the "CardEffect" as separate objects

the "Card" would be the UI object and when it is fed the backstage-object it sets its data based on the backend

and that is how most of the things work

so none of these are monobehaviors and none of them are "physically in the game" so to speak

although some of them are scriptableobjects

[undone coral]

maybe the only way to truly understand why you should make your 5 discover methods in your game manager

is to go ahead and try to finish a game where you start by writing them in a deck class

for example, you might have a monster whose behavior reacts to what card was drawn

you'll eventually achieve this by firing a DiscoveredCard event

[frozen badger]

That makes sense if it was a card game

[undone coral]

and that should illustrate a concrete example of how a Deck class would have to be invasively aware of the rest of the game

listen i don't play wow, so i can't give wow examples

lol

but you gotta work with me here

[frozen badger]

But there's no way for a monster to react to a drawn card

a monster isn't a "card" any more the second it leaves the deck

in that sense

it's only a card in the nature of "being drawn"

and being searchable, and being limited in quantity

[undone coral]

well i think you can imagine

you want some behavior in a monster

that interacts with literally anything else in the game

you'll need an event system

[frozen badger]

Like, if you want to simulate the randomness of dice, you Random.Range(0,6) right?

that gives you basically perfect dice

but if I want the randomness in cards, I can't just go Random.Range(0,52)

[undone coral]

one perspective is that 99% of these games are interesting because they innovate in the "card text"

[frozen badger]

Yeah but my game is not a card game

[undone coral]

so a monster that reads "all your dicey dungeons dice come up as even numbers now" is interesting

[frozen badger]

I don't have these cards on my hand and then play them on the field

[undone coral]

all of these games are card games. WoW mechanics are routinely prototyped as card games

because they're all spells

it's just text

it's some game rules and they're interesting

[frozen badger]

it's, when you roll a random thing, it should behave as a card

[undone coral]

i know it's not a Card Game but it is a card game

[frozen badger]

when you roll it, only

do you get what I'm trying to say?

[undone coral]

it's true for real, blizzard does prototype its games with cards

[frozen badger]

Like, the moment the rolling is done it ceases to be a card 100%

[undone coral]

like they start by describing the overwatch class as a card

i know

[frozen badger]

Sure

[undone coral]

i'm saying there's text written on the monster right?

it has some rules

i mean, it will eventually, to make a more interesting game

it casts fireball, fireball is a spell

whatever

and fireball has some text

[frozen badger]

Sure, so I have items that do things and you could call it a card if you wanted to

[undone coral]

yes

[frozen badger]

things like teleport character or heal character or whatever

[undone coral]

right exactly

[frozen badger]

but none of these things interact with the drawing of cards

[undone coral]

well they will

you might as well prepare your stuff to do that

[frozen badger]

no, because it's not a card game in that sense

[undone coral]

because 100% of your game is going to look like that

anyway that's it, that's my perspective

you can poke around the spellsource repo, literally thousands of cards, and see how card text is implemented

because it's how a lot of these games wind up looking

[frozen badger]

You ever play uh... Xcom?

[undone coral]

yes

[frozen badger]

okay, so you either hit or miss right

when you shoot

[undone coral]

xcom is definitely architected the way i am describing though

[frozen badger]

you roll 1/100 and then check against your hit chance

the randomness is dice-randomness

imagine if instead you did the randomness through cards

[undone coral]

xcom is really supporting my case

[frozen badger]

the rules for card-randomness is:
A card exists only once.
The order of cards matter.

right?

[undone coral]

for example, "immune to being critically hit (unless flanked)" is an ability

[frozen badger]

Sure, but that's unrelated to the randomness

[undone coral]

that is printed on a character

hmm

[frozen badger]

I am talking about the rolling of hit % here

which, in xcom, is dice-randomness

[undone coral]

"-60% critical chance against this unit."

that's also an ability

[frozen badger]

Right, but that's still not related to the type of randomness

[undone coral]

well like i said, this function you are worried aobut, it's very small

the random dice part is 4 lines. and by the way, it is invasive

[frozen badger]

But do you understand what I mean by random type?

[undone coral]

because i have to store the seed somewhere

[frozen badger]

Like let's say my xcom guy shoots 3 times

and I only have 1 crit in his deck

[undone coral]

and i'm not going to store it in the deck

[frozen badger]

he can only crit once

that doesn't make it a "card", but it changes how his randomness works

he can't roll crit 3 times

[undone coral]

so you would have to invasively populate the Deck class with references to a Random class instance

[frozen badger]

every not-crit increases the chance that the next one is a crit

[undone coral]

i think i finalyl gotchyu

[frozen badger]

Well it is the whole point of the class

like let's say I have a huge table of items and I just roll which item to give right

[undone coral]

yes, but it will have an instance of Random with a seed right?

i think i got you

i did it

[frozen badger]

Uh, I use the static unity random

[undone coral]

i mean you can not use a seed

[frozen badger]

which is pre-seeded

[undone coral]

well that's a mistake

all these games have seeds

that are carefully controlled

even randomness is invasive!

look i'm trying to make your life easier not harder

it turns out everything is invasive in a video game like this

[frozen badger]

yeah but in my case I mostly pull cards based on the order

right?

[undone coral]

maybe. anyway good talk

[frozen badger]

so, the "seed" is gone in that sense

[undone coral]

i gotta go

[frozen badger]

Yeah thanks for the help

[undone coral]

i think the game will be fiun

[frozen badger]

I appreciate it

I will try to show you when the randomness thing works as I want it to

[undone coral]

okay what about "this monster always gets drawn first in a playthrough"

"this monster always appears in a discover"

"until it is picked"

[frozen badger]

If there was such a thing that would be in the game mode

[undone coral]

[frozen badger]

Like the game mode would specify the starting conditions anyway

[undone coral]

it would mean some part of the random drawing would have to invasively know about the items inside the deck

okay what about

"put a monster into the deck that always appears in discover until it is drawn"

it's a monster called Corruptor and it puts a corrupt blood into your encounters pile

this is actually a really common mechanic across all these games

[frozen badger]

Yeah but you still are stuck on the word cards

Focus on dice for a sec right

[undone coral]

sure

[frozen badger]

If I roll a dice and get a 6, what is the chance that the next roll is a 6?

[undone coral]

"Next time you choose an encounter, one option is always replaced by Corrupt Blood until you choose it"

[undone coral]

yeah yeah yeah

i'm just trying to illustrate that you can call the method "PickEncounter" or whatever, and it will invasively have to know a lot about everything

your actual deck class can have the shuffle methods or whatever, it can be simple

that stuff isn't hard to write

you can move those methods if and when you need to invasively know about something when you shuffle

but it will

i already know it will

knowing nothing about your game

moving things will be easy

[frozen badger]

But you're still attached to cards as cards, not as a system for randomness

Again, consider dice

[undone coral]

sure when you roll dice

rolling dice will evnetually need to invasively know about something

okay i gotta go for real

[frozen badger]

Rolling 6 does not affect the chance of rolling a 6 again

Pulling an ace changes the chance for pulling another ace

And that behavior is the point of the deck.

[obsidian glade]

briefly read through this discussion - if the purpose here is simply to act as a random number generator with certain properties, why not use it like that and keep it simple - use a mediator that assigns values arbitrarily as needed, your generator logic only spits out a random number, and your mediator returns the object it assigned the id to

I suppose you can compress all of that into a generic Deck<T> object, but the key point is the objects you provide aren't having to supply their own ids, nor keep track of their own id

[frozen badger]

I went over to the "directly remove the object from a list by equality comparison" as suggested by doctorpangloss

so the ID itself has been removed, as has the need for an interface

[frozen badger]

This would work, but sometimes I want to get random objects by properties like "Find two random healing items" or "Find a hard encounter"

[tender light]

So what is your issue now? @frozen badger

[frozen badger]

I am doing better now, Just ended up in a somewhat winded discussion

I got rid of the ID, but kept the <T>

[obsidian glade]

yeah, didn't mean to dig up an issue again as it seems like you've already resolved it - mostly just my own curiosity in how to handle it

[tender light]

Yeah I read through it now. It looked like you had 2 completely unrelated conversations going on there near the end.

[frozen badger]

I mentioned it but I think the word "Deck" may have alluded to a card game that didn't exist and caused confusion

it was more "Deck" as in the data structure

[tender light]

I don't believe that was the issue. From my prespective doctor just tried to explain to you that manipulating a collection should be done from the outside. That the API of whatever data structure should only contain elementary operations and nothing more. Everything that has multiple effects on a particular data structure should probably not be declared as part of the API. In this case he gave multiple great example of where adding Discover method as part of a deck is a bad idea because it causes the deck to have to emit events etc... instead of it being done from some game manager type class